For the encoding and decoding of digital content, synchronization plays a vital role. For example, in a digital watermarking environment, synchronization (also called registration) is vital for payload decoding in applications such as watermarking. For audio and video watermarking, the synchronization typically takes the form of a temporal synchronization, though some video watermarking methods also require spatial synchronization. Common methods for watermark synchronization can be classified into four major categories:    [1] Geometrically invariant transform based, where the watermark is embedded in a signal transform domain that is invariant to specific geometric transformations.    [2] Recovery assisted methods, where a synchronization signal or an autocorrelation of a periodic watermark will be utilized for synchronization.    [3] Feature based, where the watermark is embedded into semantically meaningful features of the content.    [4] Exhaustive search, where the detection is running over all possible geometric distortions.
However, current synchronization schemes have noted deficiencies. For example, assume a payload, p, is an N-bits binary sequence b_0 b_1 b_2 . . . b_{N−1}, where b_i=0 or 1, for 0<=i<N. Further assume that the payload, p, will be repeatedly embedded into the content as b_0 b_1 b_2 . . . b_{N−1} b_0 b_1 b_2 . . . b_{N−1} . . . b_0 b_1 b_2 . . . b_{N−1}. During decoding, due to cropping, insertions, deletions and/or substitutions (collectively, IDS), etc., some payload bits might be lost, some might be incorrectly decoded, and some additional bits might be added. Thus the decoded sequence may look like b_3 b_4 b_6 b_7 b_8 a_0 b_9 a_1 b_11 b_12 . . . , where in this example, b_0, b_1, b_2 are cropped, b_5 is deleted, an additional bit a_0 is inserted between b_8 and b_9, b_10 is substituted by a_1.
As such, what is needed is a robust synchronization scheme for digital content that overcomes IDS degradations.